Gwern Branwen

But across deep learning in general, I think we've seen a whole range of minds and ways to think that you wouldn't find in any philosophy of mind paper.

Yeah, I'll give one example.

I was telling someone the other day –

So, you know, GAN models have incentives to hide things because it's an adversarial loss.

Whereas diffusion models have no such thing, right?

So GAN models are scared.

They put hands off the screen and they just kind of can't think about hands.

Whereas diffusion models think about hands, but they're like gigantic, monstrous, Cthulhu-esque abortions.

People weren't paying enough attention to scaling in 2020.

I'm excited by the weight loss drugs, the GLP drugs.

Their effects in general on health and addiction across all sorts of behaviors really surprised me.

No one predicted that as far as I know.

And while the results are still very preliminary, it does seem like it's real.

So I think that's going to tell us something important about human willpower and dysfunctionality.

I think it's too soon to say because we haven't actually figured out what's going on with the GLPs to even understand what they're doing at all.

What has the off target?

It's kind of crazy that activating and deactivating both work.

It's a completely crazy situation.

I don't really know what to think about the Algernon argument there.

It could be that the benefits actually decrease fitness in the fertility sense because you're going out and having a happy life instead of having kids.